UPM Institutional Repository

Assessment of near-infrared and mid-infrared spectroscopy for early detection of basal stem rot disease in oil palm plantation


Citation

Liaghat, Shohreh (2013) Assessment of near-infrared and mid-infrared spectroscopy for early detection of basal stem rot disease in oil palm plantation. ["eprint_fieldopt_thesis_type_phd" not defined] thesis, Universiti Putra Malaysia.

Abstract

Basal stem rot (BSR) is a fatal fungal (Ganoderma) disease in oil palm plantations which has a significant impact on palm oil production in Malaysia. Since there is no effective treatment to control this disease,early detection of BSR is vital for sustainable disease management. Current method of detection includes periodic visual inspection based on the symptoms of the disease which often shows up at the later stage of the disease infection and consequent laboratory analysis for confirmation. The limitations of current detection technique have led to an interest in developing alternative field-based methods that can be used for rapid diagnosis of this disease. The ultimate goal of this study was to develop an appropriate spectroscopic technique that can be used for an early and accurate detection and differentiation of Ganoderma disease with different severities. The short term goal was to evaluate the possibility of using visible (VIS) and near-infrared (NIR), and mid-infrared (MIR) spectroscopy as possible techniques for the above mentioned ultimate goal. Reflectance spectroscopy analysis ranging from visible to nearinfrared region (325-1075 nm) and mid-infrared region (2.55-25.05 μm/3921-399 cm-1) was used to analyze oil palm leaf and trunk samples of healthy (G0), mildly-infected (G1), moderately-infected (G2) and heavilyinfected (G3) trees in order to detect and quantify Ganoderma disease at different infection levels. Reflectance spectra were pre-processed and principal component analysis (PCA) was performed to obtain PC scores as input features used in different pattern recognition algorithms in order to select the best learning model of Ganoderma discrimination. Linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), k-nearest neighbor (kNN), Naïve-Bayes (NB), artificial neural networks (ANNs) and support vector machines (SVMs) classification techniques, were tested to classify the leaf and trunk samples into four levels of disease severity. The applicability of using band combinations extracted from mid-infrared spectroscopy (2.55-25.05 μm) for the detection of BSR disease in oil palm leaves was investigated using optimum index factor (OIF) and analysis of variance (ANOVA). The results indicated that LDA-based model resulted in high average overall classification accuracies of 92% (leaf samples) and 94% (trunk samples) when mid-infrared absorbance spectra were analyzed. The analysis of VIS-NIR leaf reflectance spectra, in both field and laboratory conditions, showed that kNN-based model predicted the disease with high overall average classification accuracies of 99% and 90%, respectively. Comparing the results achieved from analyzing the reflectance spectra (VIS-NIR and MIR) of leaf and trunk samples with SVM and NNclassifiers demonstrated that mid-infrared absorbance data of trunk samples with the average overall classification accuracies of 97% (standard deviation = 1%) for SVM and 97% (standard deviation = 3%)for NN resulted in better performance in classifying four classes of Ganderma infestation. Moreover, among different ratio indices resulted from band combinations method, A13.10/A9.90 could differentiate between four different classes of healthiness more accurately. Results confirmed the usefulness and efficiency of spectra-based classification approach for fast screening of BSR.


Download File

[img]
Preview
PDF
FK 2013 135RR.pdf

Download (1MB) | Preview

Additional Metadata

Item Type: Thesis (["eprint_fieldopt_thesis_type_phd" not defined])
Subject: Palm oil industry
Subject: Near infrared spectroscopy
Subject: Ganoderma diseases of plants
Call Number: FK 2013 135
Chairman Supervisor: Prof. Shattri Bin Mansor, PhD
Divisions: Faculty of Engineering
Depositing User: Haridan Mohd Jais
Date Deposited: 20 Jul 2017 10:20
Last Modified: 20 Jul 2017 10:20
URI: http://psasir.upm.edu.my/id/eprint/56200
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item